How to Use Nano Banana Pro in 2026: Official Site, Redo with Pro, and AI Studio

Nano Banana Pro does not have a separate official website, app, or login. It is Google’s name for the Gemini 3 Pro Image model. Use a Google-owned surface: Gemini for the consumer workflow, or Google AI Studio and the Gemini API for direct model selection.

The Gemini app flow changed after many early tutorials were published. As of July 20, 2026, Gemini Apps Help says ordinary image generation uses Nano Banana 2 or Nano Banana 2 Lite. A paid subscriber first creates an image, opens More, and selects Redo with Pro. Merely choosing the Pro chat model or typing “use Nano Banana Pro” does not prove that the image was generated by Pro.

Developers have a cleaner direct route: select the stable model ID gemini-3-pro-image in AI Studio or the Gemini API. Older guides that use gemini-3-pro-image-preview are out of date for a new official integration.

Start at the Official Route, Not a Lookalike Domain

Google documents Nano Banana Pro through Google products, not through a standalone Nano-Banana-branded site. Exact-match domains can provide independent tools, but their login, credits, privacy terms, models, and billing are their own contracts.

Before signing in or uploading an image, check three things:

1. The hostname is owned by Google, such as gemini.google.com, aistudio.google.com, ai.google.dev, or cloud.google.com.
2. The page identifies the model as Gemini 3 Pro Image. “Nano Banana Pro” is the product nickname, not a separate account system.
3. The surface’s terms match your task. Gemini subscriptions, Gemini API billing, and an independent image site are not interchangeable forms of access.

An independent site is not automatically fraudulent. The safe conclusion is narrower: it is not Google’s official Nano Banana Pro website, so verify its model route and data policy separately.

How to Use Nano Banana Pro in the Gemini App Now

The current consumer workflow is a quality-upgrade pass, not a hidden model toggle. You need to be signed in, meet the account and age requirements, and have an eligible Google AI plan. Availability can also vary by country, account type, and capacity.

1. Create the first image

Open Gemini, enter the image area, and create an image with the available Nano Banana 2 or Nano Banana 2 Lite path. Give the first request enough structure to produce a useful composition. Do not spend the Pro redo on a draft whose subject, aspect ratio, or content is still undecided.

2. Inspect the draft before using Pro

Check whether the image already completes the job. If the wording, hierarchy, composition, and subject are correct, downloading the Nano Banana 2 result may be the sensible stop point. Pro costs more compute and usually takes longer; it is not a mandatory final button.

Use Pro when the draft exposes a reason for a higher-fidelity pass, such as:

• a poster needs cleaner text and spacing;
• an infographic has the right logic but weak visual hierarchy;
• a product mockup needs more convincing material detail;
• a multi-subject scene needs stronger consistency;
• the final asset needs the Pro model’s higher-resolution production path.

3. Choose More → Redo with Pro

Open the image’s More menu at the lower right, then select Redo with Pro. This action is the current consumer proof that you requested the Pro pass. The model can reconstruct parts of the image, so “redo” should not be treated as a pixel-identical upscale.

Add a focused instruction if the interface allows it. Preserve what is already correct and name the defect you want fixed:

“

Keep the 16:9 composition and all three panel labels. Redo with Pro to improve text legibility, spacing, and fine material detail. Do not add new claims, badges, or people.

4. Compare, verify, and stop

Compare the first and Pro versions at full size. Keep the Pro result only if it solves the stated defect without introducing a more important error. If it changes a name, count, product shape, or factual label, the higher-detail image is not the better result.

Google’s current Help also notes that Pro redo depends on the Nano Banana 2 image quota. If that daily quota is exhausted, you cannot perform another Pro redo. Google describes limits as variable, so do not plan a production schedule around an assumed universal daily count.

A First Task That Reveals Whether Pro Helps

A generic landscape is a weak Pro test: many models can make it look attractive, and there is little to verify. A better diagnostic combines exact text, layout, relationships, and a defined canvas.

Run this as a new 16:9 image request:

“

Create a clean 16:9 event poster for “HARBOR SCIENCE NIGHT.” Put “Friday • 7:00 PM” directly below the title. Divide the lower half into three clearly separated, left-to-right panels labeled “Telescope Lab,” “Robotics Demo,” and “Ocean Map.” Use a dark navy background, warm white type, and one cyan accent. Keep all text exactly as written. Do not add sponsors, ticket prices, QR codes, or extra event details.

This is a diagnostic prompt, not a claim that we generated or validated a specific output in your account. Its value is that success is observable.

After the first image, use Redo with Pro only if you can name a defect. Then score both versions:

For a real campaign, replace the test copy with approved source text. If you upload a reference, state its role: for example, “use Image A only for the product shape; do not copy its background or text.” Google’s image-generation guide supports multiple references, but more inputs do not remove the need to check identity, count, trademarks, and permissions.

Use Nano Banana Pro Directly in Google AI Studio

AI Studio is the shortest official path when you need to choose Pro directly rather than rerun a consumer image. Open Google’s Nano Banana Pro deep link, sign in, and confirm the model selector before sending private material.

In a signed-out browser check on July 20, 2026, that link redirected to Google Account sign-in. The public AI Studio welcome page promoted the newer general-purpose Nano Banana model, so the landing-page artwork alone is not confirmation that Pro is selected. The selector must show gemini-3-pro-image after sign-in.

Use this sequence:

1. Open the official deep link and sign in to the intended Google account.
2. Confirm the selected model is gemini-3-pro-image.
3. Start at 2K unless you have a concrete reason for 4K. In the official API, 1K and 2K Pro image output share the same standard image-output price row.
4. Specify canvas, exact text, content zones, reference roles, and exclusions in the prompt.
5. Generate one result and run the same text/fact/layout checks used above.
6. Move the request into code only after the prompt and acceptance criteria are stable.

Google currently lists no free tier for Gemini 3 Pro Image API output. AI Studio access, a consumer subscription, and paid API usage are separate contracts. For the full cost rows and caveats, use the current Nano Banana Pro pricing guide.

Minimal Gemini API Example with the Stable Model ID

Create and protect a Gemini API key before running code. The dedicated Nano Banana Pro API key guide covers project, billing, and credential setup without mixing it into this consumer tutorial.

With the current google-genai Python package, a direct Pro request can be kept small:

python
import os
from google import genai
from google.genai import types

client = genai.Client(api_key=os.environ["GEMINI_API_KEY"])

response = client.models.generate_content(
    model="gemini-3-pro-image",
    contents=[
        "Create a clean 16:9 technical diagram with the exact title "
        "'IMAGE REVIEW FLOW' and three labeled stages: Draft, Verify, Publish."
    ],
    config=types.GenerateContentConfig(
        response_modalities=["TEXT", "IMAGE"],
        image_config=types.ImageConfig(
            aspect_ratio="16:9",
            image_size="2K",
        ),
    ),
)

for part in response.candidates[0].content.parts:
    image = part.as_image()
    if image is not None:
        image.save("image-review-flow.png")

Three details prevent common setup mistakes:

• Use the stable gemini-3-pro-image, not the older preview ID.
• Keep the K uppercase in 1K, 2K, and 4K.
• Treat successful generation as the start of review, not proof that every rendered word or diagram fact is correct.

For current aspect ratios, reference-image limits, response handling, and SDK variants, use Google’s official image-generation documentation as the implementation owner.

Why You Cannot Find Redo with Pro

Do not keep clicking model selectors or follow “hidden activation” instructions. Work through the failure class instead.

If the failure is an API error rather than access confusion, use the Gemini image common-errors guide or the focused Nano Banana Pro error-code guide. Repeatedly changing prompts will not repair authentication, billing, region, or model-name failures.

When Not to Use Nano Banana Pro

Pro is the wrong default when its strengths are not the bottleneck.

Choose Nano Banana 2 first when you need rapid ideation, many drafts, everyday photo edits, or a cost-and-latency-balanced API workflow. Choose Pro when a final asset depends on precise text, a complex layout, high-fidelity product detail, localization, multiple controlled references, or 4K output.

The decision rule is simple:

• Use the Gemini app redo when a good consumer draft needs one higher-fidelity final pass.
• Use AI Studio/API Pro directly when model control, resolution, repeatability, or integration is part of the job.
• Stay with Nano Banana 2 when the draft already passes, the work is exploratory, or iteration speed matters more than final detail.
• Stop and use a conventional design tool or human review when exact legal copy, regulated facts, pixel-locked branding, or error-free diagrams are mandatory.

Google DeepMind’s model limitations explicitly warn that spelling, fine detail, factual data, translation, complex edits, image blending, and character consistency can still fail. Pro reduces some failure rates; it does not convert generated output into verified source material.

For a workload-by-workload comparison, see Nano Banana Pro vs. Nano Banana 2. If the route is settled and prompt quality is now the constraint, continue with the practical Nano Banana prompting guide.

Common Questions

Is Nano Banana Pro free or unlimited?

Do not rely on either word without naming the surface. Current Gemini Help makes Pro redo a paid-plan feature and says limits may change. The official Gemini 3 Pro Image API pricing page lists no free API tier. No Google source supports a universal “free unlimited Nano Banana Pro” promise.

Is Nano Banana Pro an app by Google?

It is a Google image model used inside Google products, not a standalone Nano Banana Pro app. The official consumer app is Gemini. The official developer surfaces are Google AI Studio, Gemini API documentation, and Vertex AI.

Can Nano Banana Pro make videos?

Gemini 3 Pro Image generates and edits images. It can make a storyboard, keyframe, poster, or visual asset for a video workflow, but it is not the video-generation model. Do not interpret motion-looking imagery or a third-party product bundle as proof of native video output.

How do I get the best result?

Define the canvas, exact text, information zones, reference roles, and exclusions before generation. Then change one meaningful defect at a time and compare against a written acceptance checklist. More prompt adjectives are not a substitute for source accuracy or design review.

The Reliable 2026 Workflow

The current Nano Banana Pro workflow is easier once the routes are separated:

1. Use Google-owned domains and identify the model as Gemini 3 Pro Image.
2. In Gemini, generate first and use More → Redo with Pro only on an eligible paid account.
3. In AI Studio or code, confirm the stable ID gemini-3-pro-image before submitting the prompt.
4. Give Pro a task with measurable requirements, not just an aesthetic theme.
5. Verify text, facts, counts, layout, identity, and output size before publishing.
6. Stop when Nano Banana 2 already passes, access prerequisites are missing, or generated content cannot meet the required accuracy boundary.

That route-and-proof approach is more durable than memorizing an old button sequence. It also answers the question that matters after the image appears: not merely whether Nano Banana Pro ran, but whether the result is safe and useful for the job.